How to limit voltage extremes on PWM

T Chap

I am stuck on how to do this. The obvious first thought was to use a trim pot as shown in the drawing, I put the trim pot in the drawing just for a visual. The idea is, I want to be able to set a maximum volume level on a line level audio path using the Opto Jfet to pull the level down, this serves several purposes, as the fet can be ramped up or down to prevent pops, hard switching is too abrupt for the application and I want programmable ramp times. So the goal is to use a trim pot (or 8 bit ladder) as a master volume that controls several jfets at once, using a scheme where the input to the LED has a lower level limit, meaning that at 3v3 the audio if fully muted as the jfet is effectively shorting the audio path the GND before it sees an audio amp down the line. When the LED is seeing 0vdc, then the jfet is doing nothing to the audio, the audio is at max volume. I want to limit the max volume by setting a limit to the lower level voltage derived from either PWM or logic low so it never hits 0, but can sit where it needs for the project. This way, there is a single fixed lower level limiter that can affect multiple fets.

The trim pot is the right idea for a starting point, but it is not ideal, as there is loss of current, but the trim pot does seem to work as long as you park the wiper at one extreme or the other. Can anyone suggest a better way to accomplish limiting the lower extreme of PWM or logic 0 without affecting the high side(3v3)?

Peter Jakacki

I can't see why you bothered with a 4049 as the Prop's I/O is quite happy to drive a simple RC filter directly plus your configuration doesn't seem right. However the output of the PWM RC filter is a voltage and needs to be buffered if you want it to supply any current for the LED. Putting that aside though all you need to do is to couple the Prop's I/O directly to the LED through the 180R resistor and forget about the RC filter as you want the full voltage to forward bias the infrared led and as long as your PWM frequency is high enough the PWM frequency can be filtered out.

This may not be the case however and so what you really need to do then is run the Prop to an RC filter and then buffer it with a simple NPN emitter follower to supply the led current. You will not get the full range of PWM though because anything below the led forward voltage drop (around 1.2V) will not work. Also, the response of the led and JFET combo may not be linear.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
*Peter*

T Chap

Peter The 4049 serves several purposes, one is that there are two of the Optos that have to be turned on off one pin from the Prop so I mult the Prop output to two gates. The other is that with the inverter, on power up the audio outputs are shorted to gnd and you don't hear the VS1033 power up with its pop. I was using the 4049 gates to provide the current to the two LEDs. I will look at the NPN idea now to see if it will solve the remaining issue of how to set a threshold so the LED input can stay above GND at some preset level.

Peter Jakacki

Thanks for the clarification Todd, that makes sense now that you say you want it to mute on power-up. BTW, the Prop I/O has no problems driving this in an open-drain configuration really, you just toggle the direction bit rather than the I/O for this.

Since the led is a current device you could drive in a fashion that overcomes the forward bias required. Using the NPN as an emitter follower but placing the led in series with the collector and supply you effectively have created a voltage to current converter (and buffer). The current is set simply as V/R where V is the PWM voltage less 0.6V base-emitter drop and R is the emitter resistor. Please note that you do not use the 180R resistor at all as the NPN+R becomes the constant current source. For 20ma full current you would say that R=V/I where V = 3.3-0.6 = 2.7 so 2.7/0.02 = 135R so a value of 120R would be good for starters.

EDIT: The base-emitter has some resistance too but 120R or even 100R is still good for starters.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
*Peter*

Post Edited (Peter Jakacki) : 6/6/2010 1:12:09 AM GMT

T Chap

Thanks Peter, I will try the idea out.

Beau Schwabe

Todd Chapman,

"... meaning that at 3v3 the audio if fully muted as the jfet is effectively shorting the audio path the GND before it sees an audio amp down the line. When the LED is seeing 0vdc, then the jfet is doing nothing to the audio, the audio is at max volume ..."

You sure?

a J-fet should be 'ON' without power applied. When you apply power is when it turns 'OFF'.

If your device turns 'ON' with power, then it is a MOSFET, in which case you need two of them in series back -to-back to negate the internal reverse diode.


back-to-back implies that both Drains are tied together so that one Source goes to GND and the other Source goes to the 1K resistor as you have it in one of your other schematics.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Beau Schwabe

IC Layout Engineer
Parallax, Inc.

Post Edited (Beau Schwabe (Parallax)) : 6/6/2010 4:54:40 AM GMT

Peter KG6LSE

Beau Schwabe (Parallax) said...

If your device turns 'ON' with power, then it is a MOSFET, in which case you need two of them in series back -to-back to negate the internal reverse diode.

to add to Beau: there is 2 kinds of MOSFETS ...

Enhancement mode and Depletion mode .
both are opposite of each other [noparse]:)[/noparse]

Peter KG6LSE

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"Carpe Ducktum" "seize the tape!!"
peterthethinker.com/tesla/Venom/Venom.html
Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway. —Tanenbaum, Andrew S.
LOL

T Chap

I made a lot of progress, almost done.

I thought it was a Jfet, but is a Photo FET H11F1M When you turn on the LED, it acts just like a variable resistor and drops to about 100ohm, basically open with no voltage. I will post the working schematic later.

Post Edited (Todd Chapman) : 6/6/2010 6:38:36 AM GMT

T Chap

Thanks Peter for the ideas on the NPN, that really helped out a lot. Using one Prop pin straight to 2 3904's, the Prop can set the volume for two audio lines. What is really great is that I found a convoluted method to use counters in the background to set and forget the volume levels with a ton of resolution between full on and full off, plus fade in/out the audio when needed to work around some inherent audio glitches when starting and stopping mp3's on a Vinculum/VS1033 combo.

This could be use for a crude DA with no cogs or resources required to set a voltage anywhere you want and leave it alone after it is set. The nice thing about the 8 counters on the Prop are that you have 8 DAC's that can run behind the scenes forever until you change them.

CON
  _clkmode               = xtal1 + pll16x
  _xinfreq               = 5_000_000


obj

VAR
  long rate, x, y, volmax

PUB Start
  Y := 1_000_000_000   'start muted
  DAC_START(13, y)
  dira[noparse][[/noparse]13] := 1

  SetVol(100)  '% max vol  0 = off  100 = max
  'testing
  repeat
    'FadeIn
    'waitcnt(80_000_000 + cnt)
    'SetVol(10)
    'waitcnt(80_000_000 + cnt)
    'FadeOut
    'SetVol(100)
    'waitcnt(80_000_000 + cnt)

PUB SetVol(mult)
    Y := 1_000_000_0  * (100 - mult)
    DAC(y)


PUB FadeIn
   volmax :=  1_000_000_00
   Y := 1_000_000_000
   rate := 50000
   DAC_START(13, y)
   dira[noparse][[/noparse]13] := 1
   DAC(y)
   REPEAT while Y > volmax
     Y :=  Y - rate
     DAC(y)
     waitcnt(40_00 + cnt)
   'REPEAT


PUB FadeOut
   volmax :=  1_000_000_000
   'Y := 1_000_000_00
   rate := 20000
   REPEAT while Y < volmax
     Y :=  Y + rate
     DAC(y)
     waitcnt(40_00 + cnt)
   'REPEAT



PUB DAC_START(PIN, freq)
    ctra := %00110 << 26 +  pin

PUB DAC(Value)
    frqa := value

PUB DAC_STOP(PIN)
    ctra := %00000 << 26' +  pin

Post Edited (Todd Chapman) : 6/6/2010 8:34:29 AM GMT

Peter Jakacki

I didn't worry when you said it was an opto jfet as I figured it was one of those typical optofet devices that are designed for AC signals (having a part# helps). I looked at your schematic but it's missing the resistor from the emitter to Vss. This will sort of work but not as well as having that resistor in the emitter. Without a resistor the PWM voltage will always be clamped at 0.6V plus it is possible to fully turn on the NPN and burn out the opto's led.

If your control voltage is common to both channels you can put both optos in series using a single NPN. I have attached a sample schematic to show how I might hook it up. Notice that the RC circuit is driven directly from the Prop and that there is a separate 10K pullup which makes sure that the opto is fully on when the I/O is still not set.

Yes, using duty-mode output from the counters (2 per cog = 16) is a better way of generating a voltage.

TIP: Do you know that if you use an amp like the TDA7052 in place of the LM386 that you can get more power at +5V into the speakers and without all those bulky coupling and compensation capacitors.

EDIT: returning the optos to a higher voltage than VDD would be beneficial as the current source would drop 2.7V max and the optos around 1.2V each so even +5V would be good.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
*Peter*

Post Edited (Peter Jakacki) : 6/6/2010 9:18:35 AM GMT

Phil Pilgrim (PhiPi)

Peter, your circuit is an improvement, and I especially like the emitter resistor to limit IRED current. But two things come to mind:

1. The IRED's V_fwd(max) is 1.75V. If Vdd is 3.3V, it's not high enough to have the IREDs in series. (Even driving them from 5V is not enough, given the 1.6V emitter voltage at 16mA.).)

2. The base and pullup resistor values are too high to drive the transistor adequately. For an IRED current of 16mA, the emitter resistor must have 1.6V on it, which means the base has to be driven at 2.2V. With a 3.3V drive through the two 10K resistors (on reset), that's only 55uA of base current. The transistor would have to have a minimum current gain of about 300 for this to be effective, which is much higher than that of the 2N3904.

I'd be more incliined to use the isolator in series mode to pass the audio, rather than shunt mode to choke it. That way the audio will pass only when the IREDs are being driven, precluding any special treatment when the Prop pin is not being driven. Also, the optos' switching time is around 25usec, which is pretty slow. So you may not even need the cap, if you use a DUTY mode output to drive the transistor.

-Phil

T Chap

Thanks guys for the great info. I will test all these ideas today and see how it goes. I am not apposed to the TDA7052, just don't have one but will put it in for the board build. Yesterday I could not get the LED hot enough to fully shut off the audio (shunt mode) with even a 50ohm off the emitter, so I felt like the LED had more to give and didn't sweat burning it out, although it would appear that it should have it naturally. Without the 1uF there the behavior is not good on the audio, it has an aliasing noise that is pretty bad. There is easy access to 5v so I can test with that and put back in the series resistor and see how it lights up.

Running the audio with the isolator in series is doable, I tried lots of combinations of shunt and series, and spent all night just trying to solve the aliasing noise which was present both modes until I changed from .1 to 1uF, and just happened to be in shunt when it got resolved. I am using Duty and find that there are strange noise artifacts without it. Testing NCO there was no frequency that worked for audio. Keep in mind that I am not using any methods to keep the phsa resetting as would be required for PWM, basically just setting the clock freq and letting it run, unless it is time to ramp it up or down, then a method handles that and parks it again.

Using 5v then the series resister to GND and to the base are fine.

Post Edited (Todd Chapman) : 6/6/2010 6:07:01 PM GMT

Peter Jakacki

Phil, yeah the values are basically placeholders I suppose, the main point was to clear up the errors in the Todd's schematics. If the opto is that slow it could be driven with PWM directly from the Prop through a current limit resistor. Duty mode would be too high a frequency to work effectively but a suitable frequency PWM could do it. The typical forward votlage drop of the led is 1.3V @16ma though and using the optos in series really needs a higher supply BUT +5V should do it although it really only saves a couple of components. Suggestions suggestions. I would have to have a good look at the opto's specs to really say much more though.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
*Peter*

T Chap

Peter and Phil

I really appreciate the detailed responses. I went through all the notes today and have it working really nicely. I don't have the TDA7052a amps yet but tested on several hi fi systems. These optofets sure make for a cheap method to control audio without having to have dual supplies and other expensive parts. Below are the final drawings and test code. The base has a weak pullup for poweron which keeps things muted until something else takes over. The audio did not sound right using the fet in series (consumer CD player line out). The shunt works well with a 100k resister in series and shorting the amp side, without the series R the audio sounds bad between the extremes of full on and full off, in between it gets distorted for some unknown reason.

CON
  _clkmode               = xtal1 + pll16x
  _xinfreq               = 5_000_000

VAR
  long rate, x, y, volmax

PUB Start
  Y := 2_147_000_000   'start muted 2_147_000_000 = MUTED  0 = Unmuted
  DAC_START(13, y)
  dira[noparse][[/noparse]13] := 1
  repeat   'tests
    FadeIn
    waitcnt(80_000_000 + cnt)
    FadeOut
    waitcnt(80_000_000 + cnt)

PUB SetVol(val)    '2_147_000_000 = MUTED  0 = Unmuted
    Y := val
    DAC(y)


PUB FadeIn
   volmax :=  0
   Y := 2_147_000_000
   rate := 50000
   REPEAT while Y > volmax
     Y :=  Y - rate
     DAC(y)
     waitcnt(40_0 + cnt)

PUB FadeOut
   volmax :=  2_147_000_000
   Y := 0
   rate := 30000
   REPEAT while Y < volmax
     Y :=  Y + rate
     DAC(y)
     waitcnt(40_00 + cnt)

PUB DAC_START(PIN, freq)
    ctra := %00110 << 26 +  pin
    frqa := freq

PUB DAC(Value)
    frqa := value

PUB DAC_STOP(PIN)
    ctra := %00000 << 26' +  pin

Post Edited (Todd Chapman) : 6/7/2010 4:47:44 AM GMT

Peter Jakacki

Ok, for some reason your revised schematics never seem to be correct as if you jump the gun or reinterpret things. The RC network from the Prop should be PROP---->R---->C(gnd)--->NPN.base. But you are getting there.

The 1uf appears as a dead-short to anything but DC from the Prop plus it will discharge as quickly as it charges, not what you want. Use the 1K from the Prop to charge the capacitor which BTW can now be a much smaller value. The output of this RC filter is fed straight into the base of the NPN without any other current limit resistors required as the input resistance of this is around gain*(Re+100) which is at least 12.5K.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
*Peter*

Phil Pilgrim (PhiPi)

Todd,

What happens if you remove C5? The reason I'm asking is that it's probably not serving any purpose without some series resistance from the Prop pin. Moreover, driving a cap that large without the series resistance could result in pin failure, due to the large switching currents.

-Phil

T Chap

I just got rid of the 1uF and put a .1 at the base to GND. It doesn't need the large cap anymore, I think it was helping prior to adding the transistor. Thanks for the tips!

Post Edited (Todd Chapman) : 6/7/2010 6:12:10 AM GMT

T Chap

Peter, on the TDA7052A/AT do you know if it is OK to short the speaker - to GND?

Peter Jakacki

That's one of those trick questions, isn't it Todd? I'm wise to that....yeah.

Come on, why would you want to short it to ground? The whole idea of a BTL (bridge tied load) amp is to drive the speaker differentially so you need 2 outputs without a ground. If you were to try to drive the speaker single ended then you would need a big capacitor to couple the AC and block the large DC that would otherwise flow but that's what the LM386 does already.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
*Peter*

T Chap

Ha. I assumed the answer but I wanted it to be otherwise [noparse]:)[/noparse]

Peter Jakacki

Ok, I'm taking a guess at why you asked since you didn't volunteer.

The BTL amps are best for driving speakers but line outputs need to be single-ended though low power and headphones usually have a common but they can be driven by a much simpler stereo headphone amp.

EDIT: You should be able to drive each side of the headphones with the single ended output of the BTL amp but coupled through a small capacitor. Should work.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
*Peter*

T Chap

Thanks Peter, I will leave the IC in as it has great reviews versus 386 plus is cheaper with lower associate part count, but it required a good bit of work down the line to fix the issue of no longer having common GND for the outputs.